search

ML4GW / aframe

Detecting binary black hole mergers in LIGO with neural networks
MIT License
18 stars 17 forks source link

Fixed shift calculation #396

Closed wbenoit26 closed 1 year ago

wbenoit26 commented 1 year ago

Closes #394

Fixes the bug in the shift calculation that sometimes led to an extra shift, and updates it to reflect not using the zero-lag data.

EthanMarx commented 1 year ago

It looks like the correct calc was implemented in the timeslide_waveforms project however. I would expect the infer project to break since there wouldn't be any waveforms to inject. Is there an explanation?

wbenoit26 commented 1 year ago

...Huh. That's interesting. No, I don't have an explanation for that, but I can look into it.

wbenoit26 commented 1 year ago

Just confirming, for 5 shifts of a background of 1,600,000 seconds, we could have fit ~150,000 waveforms. Instead, we have ~120,000 waveforms, which makes sense for 4 shifts. So the correct number of waveforms were being generated, but somehow that doesn't break inference.

wbenoit26 commented 1 year ago

@EthanMarx I think it makes sense just to get this in, rather than generalize things so function isn't in two places. I'll make an issue for it though.

EthanMarx commented 1 year ago

Yeah I agree, sounds good